Advanced searches left 3/3

1 Bit - Astrophysics Data System

Summarized by Plex Scholar
Last Updated: 03 May 2022

* If you want to update the article please login/register

Strongest Angle-of-Arrival Estimation for Hybrid Millimeter Wave Architecture with 1-Bit A/D Equipped at Transceivers

This paper describes the most effective angles of arrival estimation for a hybrid millimeter wave communications device with 1-bit analog-to-digital/digital/analog converters embedded at transceivers. The proposed algorithm is designed to minimize the needed number of estimation overheads while keeping the root mean square error of the best AoA estimates at the base station. Simulation results demonstrate the efficiency of our proposed algorithm in terms of reducing estimation overheads while still maintaining good estimation results in low SNRs.

Source link: https://ui.adsabs.harvard.edu/abs/2022Senso..22.3140L/abstract


SensiMix: Sensitivity-Aware 8-bit index & 1-bit value mixed precision quantization for BERT compression

A Pre-training language model, such as BERT, is extremely helpful in improving the results of natural language processing tasks. SensiMix effectively uses 8-bit index quantization and 1-bit value quantization to the critical and insecure areas of BERT, maximizing the compression rate while minimizing the accuracy drop. SensiMix converts the original BERT model to a smaller, lighter version, reducing the model size by a factor of 8 and lowering the inference time by around 80% without noticeable accuracy decrease.

Source link: https://ui.adsabs.harvard.edu/abs/2022PLoSO..1765621P/abstract


Deep learning for 1-bit compressed sensing-based superimposed CSI feedback

Multiple-input multiple-output devices, 1-bit compressed sensing, and large processing delays have all shown many benefits, but several limitations remain to be overcome, such as poor accuracy of the downlink CSI recovery and lengthy processing delays. The downlink CSI is compressed with the 1-bit CS method, which is superimposed on the uplink user data sequences, and then sent back to the base station, which is compressed by the 1-bit CS scheme. With the simple traditional procedure and a single hidden layer network, Then, a lightweight reconstruction procedure that uses an initial feature extraction of the downlink CSI with the simplified traditional approach and a single hidden layer network, is used to reconstruct the downlink CSI with low processing delay. The new strategy, compared to the 1-bit CS-based superimposed CSI feedback scheme, increases the recovery time of the UL-US and downlink CSI with reduced processing time and robustness against parameter variations.

Source link: https://ui.adsabs.harvard.edu/abs/2022PLoSO..1765109Q/abstract


Online RIS Configuration Learning for Arbitrary Large Numbers of $1$-Bit Phase Resolution Elements

We discuss RISs with 1-bit phase resolution elements in this paper, and we model each of them's behavior as a binary vector with the most realistic reflection coefficients. We then introduced two new Deep Q-Network and Deep Deterministic Policy Gradient agents, aimed at a more effective investigation of the binary action spaces. In addition, the results of the new strategy is similar to the proposed learning approach when dealing with modest scale RIS sizes, where the traditional DQN based on configuration-based action spaces is possible.

Source link: https://ui.adsabs.harvard.edu/abs/2022arXiv220408367S/abstract

* Please keep in mind that all text is summarized by machine, we do not bear any responsibility, and you should always check original source before taking any actions

* Please keep in mind that all text is summarized by machine, we do not bear any responsibility, and you should always check original source before taking any actions